IEEE to Study The Feasibility of 400 GB/s Ethernet

[RedundantInk]

The IEEE has launched the IEEE 802.3 "Standard for Ethernet" study group to explore the development of a 400 GB/s Ethernet standard.

IEEE to Study The Feasibility of 400 GB/s Ethernet : Read more

 

[Hupiscratch]

Yay, UHD videos over the LAN!

 

[Dupontrocks11]

It would be cool if they reached such high speeds throught ethernet that they could send hugh density data through the intrawebs. Imagine encoding and decoding DNA samples through the web!

 

[wanderer11]

400 GB/s or 400 Gb/s? Seems you guys capitalized the b on accident.

 

[nforce4max]

Would be nice to have that in the US if it were not for those WW2 era coax and decades old copper that is still in use (hinders bandwidth). The other is providers over selling their connections while promising unrealistic speeds. So a connection like this would be fresh air to many ordinary consumers and businesses.

 

[slomo4sho]

Now only if the US ISPs could keep up.

 

[PudgyChicken]

If only our storage mediums could read/write at those speeds...

 

[frontliner]

My biggest pet peeve is messing up B & b. In this case do you mean 400GB (3,200Gb) or 400Gb? An 8x error is something to scoff at...

 

[jhansonxi]

Expect another round of half-baked incompatible standards, IP lawsuits, and "preliminary" standard implementations before components from different vendors work together properly. Hopefully it won't be as bad as the wireless standards.

 

[cravin]

ermahgerd

 

[TheViper]

...capacity requirements of 1 Tb/s in 2014 and 10 Tb/s in 2020.

...the next critical step in evolving the IEEE 802.3 standard to stay ahead of industry's needs.

Am I missing something? How is 400 Gb/s staying ahead of a 1 Tb/s 2014 industry need? Much less 10 Tb's in 2020.

 

[game junky]

This will help to kill a bandwidth bottleneck within an internal network, but until ISPs can step up to Google's challenge and create high-speed networks pushing the GB download/upload then it isn't really going to see much benefit outside of office environments. I hope Goggle's thrown gauntlet will force major ISPs like Time Warner, Verizon, AT&T & Cox to cowboy up and put new infrasture in every metropolitan area. It isn't that they're incabable, it's that they haven't been forced to make an effort to improve performance. I wonder how painful the KC-Metro experience has been for other providers in the area - makes me want to move to Olathe/Lenexa just to enjoy that sweet, sweet download speed. When cellular internet has caught up to cable broadband, it's time to get working.

In short, I am all for new ethernet standards but I am more interested to see ISPs either pushing the limit of the metropolitan fiber rings or finding another way to increase speeds to their customers without emptying my wallet.

 

[InvalidError]

Seems a little "late" to talk about 400Gbps since Google and Facebook have already been looking for Tbs-scale solutions for a couple of years already. The IEEE should really be talking about 1/4/10Tbps.

But how much is there to talk about? Simply extend what they have already been doing for 100GbE (4x25Gbps either as discrete fibers or WDM) to 1Tbps (ex.: 25x 40Gbps DWDM) and call it a day... at that sort of rate and density, the pluggable optics would be large, the interface would be pretty wide, quite power-hungry and the line-card itself would consume multiple slots worth of bandwidth in today's chassis so the exact details of how the fiber side meets the line card/backplane are of relatively little importance... whatever pluggable module spec they might end up publishing, they would end up having to periodically review to accommodate smaller variants as components and power requirements shrink.

So the logical thing to do would be to publish the optical signaling specs to ensure compatibility with future QFP++ modules and let manufacturers deal with how they want to implement the optical-to-card interface until components have shrunk down to manageable sizes that allow fitting 4+ ports per card.

 

[InvalidError]

[citation][nom]game junky[/nom]This will help to kill a bandwidth bottleneck within an internal network, but until ISPs can step up to Google's challenge and create high-speed networks pushing the GB download/upload then it isn't really going to see much benefit outside of office environments.[/citation]
Not many internal networks use more than 10Gbps anywhere near end-users unless your office computers are connected directly to a switch/router that handles substantially more intensive traffic than typical office use.

40+Gbps is mostly used for WAN/MAN/transit networks and telco/cableco/datacenter backbones and major NNIs... we are way beyond the office there. With 100s of Gbps at play there, this is where 400+GbE would be most useful.

 

[anti-painkilla]

[citation][nom]InvalidError[/nom]Not many internal networks use more than 10Gbps anywhere near end-users unless your office computers are connected directly to a switch/router that handles substantially more intensive traffic than typical office use.40+Gbps is mostly used for WAN/MAN/transit networks and telco/cableco/datacenter backbones and major NNIs... we are way beyond the office there. With 100s of Gbps at play there, this is where 400+GbE would be most useful.[/citation]

Agreed, a lot of companies would still be using Gigabit core switches. Awesome that they are looking into higher Ethernet standards though.

 

[tim1935]

I wonder what TPI you'll need for a cable like that?

 

[InvalidError]

[citation][nom]tim1935[/nom]I wonder what TPI you'll need for a cable like that?[/citation]
Zero.

The IEEE did not initially intend for there to be an electrical version of 100GbE but they whipped up the 100GBase-CR10 spec just in case someone had the funny idea of hooking that directly to a server but this is not expected to become a common practice any time soon.

In all likelihood, there will never be an electrical spec for 400GbE and beyond.